The DHS Program User Forum      
Discussions regarding The DHS Program data and results
Home » Data » Merging data files » creating panel of EAs?
creating panel of EAs? [message #13860] Mon, 15 January 2018 13:04 Go to next message
malaria2017 is currently offline  malaria2017
Messages: 1
Registered: January 2018
Location: UK
Member
I would like to know to what extent it is possible to create a panel of DHS clusters under the following circumstances: if two separate waves of DHS surveys in a country are using the same sampling frame (eg the country census of year Y), my understanding is that each DHS wave would usually draw a random sample of clusters from the same list of Enumeration Areas (EAs) based on that sampling frame.
Then in the DHS dataset, each of the clusters dranw has a unique value identified in variable v001.
Now, if a given EA is drawn in both survey waves (both using the same sampling frame), would the same value of v001 identify the same cluster??? Or would the value of v001 be completely meaningless, and the same value could indicate completely different EAs from the original sampling frame?

The reason I am asking is that UI have seen it done in a paper presented at a conference, and I am wondering how they did it?..
Re: creating panel of EAs? [message #13863 is a reply to message #13860] Mon, 15 January 2018 17:05 Go to previous messageGo to next message
Reduced-For(u)m
Messages: 290
Registered: March 2013
Senior Member

My understanding is that generating such a panel is impossible. I have also seen it done and then asked how that was possible, and the general impression I get is that people just assumed V001==1 means the same thing in each survey round - I am about 99% certain that it does not.

There may be some possibilities for generating panels at other levels, such as region (once they are coded to comparable regions, which change across surveys), and there may be ways to use the GPS to guess at clusters that were close together (even though there is random displacement in GPS coordinates) but in general I think that trying to make the cluster-level panel is a fundamental mistake people keep making.

I would love it if one of the DHS folks could confirm all this, but I've definitely noticed the same thing and been very, very suspicious.
Re: creating panel of EAs? [message #14289 is a reply to message #13863] Mon, 19 March 2018 09:03 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 1362
Registered: February 2013
Senior Member
Following is a response from Senior DHS Stata Specialist, Tom Pullum:

The clusters (enumeration areas) in successive surveys are different. The codes assigned (e.g. 1, 2,....) have nothing to do with the codes or numbers or labels in the sampling frame. They are not even in the same sequence. If anyone is actually assuming continuity from one survey to the next, they are making a serious mistake.
Previous Topic: Merging and Weighting PR and KR
Next Topic: Merging DHS data sets with municipality information
Goto Forum:
  


Current Time: Mon May 21 22:28:30 Eastern Daylight Time 2018